Functional Network-Based Analysis of Industrial Robot Dynamics

Industrial robots are attractive for automating complex manufacturing tasks because they offer high flexibility and a favorable ratio between installation and reachable workspace. However, their performance is constrained by interacting error sources arising from serial kinematics and gear drives, including compliance, friction, and backlash, whose combined effects are difficult to capture with conventional single-effect models. This study investigates whether a network-based analysis for time series can reveal how mechanical and control variables interact during motion. Functional networks provide an unconventional approach to studying interrelationships between measurements from the different variables. Each measurement time series is represented as a node in the network, and edges are inferred from the dynamics using cross-recurrent measures. In this work, internal measurements of an axis of an industrial robot, that includes encoder positions, torques, and currents, are recorded during controlled experiments. From these signals, functional networks are constructed to describe the coupling structure between variables under varying dynamic conditions. The analysis shows that changes in the velocity and the backlash effect alter the strength and symmetry of these couplings that reflects different dynamic behavior of the system. The results indicate that network-based representations can provide a complementary perspective for analyzing motion and understanding dynamic interactions in industrial robots.